## wp17217

## Source details

**Canonical URL:** [wp17217](https://www.imf.org/-/media/files/publications/wp/2017/wp17217.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp17217.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp17217.pdf.json)

---

### Introduction and research objective
- Focus: Assess macroeconomic effects of public investment in emerging markets and low-income countries using exogenous identification of unanticipated public investment shocks.
- Sample: unbalanced panel of 79 emerging market and developing economies over the period 1990-2013.
- Identification approach chosen to mitigate “fiscal foresight”: Auerbach and Gorodnichenko (AG) forecast-error approach using IMF World Economic Outlook October forecasts.

### Identification strategy and rationale
- Public investment forecast-error (FE) construction:
  - FE_{i,t} = ∆lnPI_{i,t} − ∆lnPI^E_{i,t} = (lnPI_{i,t} − lnPI_{i,t−1}) − (lnPI^E_{i,t} − lnPI_{i,t−1}).
- Rationale for October forecast choice: minimizes likelihood that unanticipated within-quarter business cycle news drive FE and aligns agents’ and econometrician’s information sets.
- Distributional features of FE shocks:
  - Range: -195 to 185 percent.
  - 90 percent of shocks are between -45 and 43 percent.
  - Average (median) shock: about -1.8 (-1.7) percent.
  - Actual growth rate of public investment in full sample ranges from -186 to 7586.

### Empirical method
- Estimation: Local projection (Jordà, 2005) to estimate impulse-response functions for horizons k = 0, . . , 5 (k = 0 is year of shock).
- Baseline specification (equation (2)):
  - y_{i,t+k} − y_{i,t−1} = α_i^k + θ_t^k + β^k FE_{i,t} + θ^k X_{i,t} + ε_{i,t}^k
  - y is log output; α_i are country fixed effects; θ_t are time fixed effects; X_{i,t} includes two lags of FE and two lags of GDP growth.
  - Confidence bands via clustered robust standard errors at the country level.
- Nonlinear specification allowing state-dependence (equation (3)):
  - y_{i,t+k} − y_{i,t−1} = α_i^k + θ_t^k + β1^k G(z_{it}) FE_{i,t} + β2^k (1−G(z_{it})) FE_{i,t} + θ^k M_{i,t} + ε_{i,t,k}
  - G(z_{it}) = exp(−γ z_{it}) / (1 + exp(−γ z_{it})), γ > 0; γ set to 1.5 (results robust to alternative γ).
  - z: business cycle indicator (real GDP growth) or country-specific characteristic normalized to zero mean and unit variance.
  - Public Investment Management Index (PIMI) used as measure of investment efficiency.

### Key empirical findings and quantitative results
- Average effects on output and employment:
  - Public investment increases, on average, output in the short and medium term.
  - Estimated short-term fiscal multiplier for developing economies: about 0.2.
  - A 10 percent unanticipated shock to government investment spending:
    - Short-term effect on the level of real output: 0.1 percent in the same year.
    - Five years after the shock, the level of real output is about 0.4 percent higher.
    - Short-term increase in employment: about 0.15 percent.
- Investment crowding and composition:
  - Public investment shocks, on average, tend to crowd-in private investment.
  - Short-term fiscal multiplier for public consumption: about 0.1 (computed as the ratio of the estimated coefficient to the public consumption-output share), slightly smaller than for public investment (0.2).
  - Public investment delivers larger medium-term output effects than government consumption.
- Distributional outcome:
  - A 10 percent increase in public investment is associated with a reduction in the Gini coefficient of about 0.2 percent (approximately ¾ of the standard deviation of the average annual change in the Gini coefficient in the sample).

### Heterogeneity of effects across episodes and country characteristics
- Business cycle:
  - Response of output tends to be higher during periods of recessions than in expansions; differences small relative to standard errors and mostly not statistically significant. Statistically significant only 4 years after the shock.
- Exchange rate regime:
  - Effects substantially stronger in countries with fixed exchange rate regimes: increase in GDP positive under fixed regimes, negative (and insignificant) under flexible regimes. Difference statistically significant only for k=0 and k=4 given large standard errors.
- Trade openness:
  - Output response positive in countries with very low trade openness, negative (and insignificant) in countries with very high trade openness. Differences mostly not statistically significant.
- Debt-to-GDP:
  - Effects vary significantly with initial public debt:
    - Countries with very low debt: response of output positive, significant and higher than baseline.
    - Countries with very high debt: response negative and not statistically significant.
  - Difference in impulse response functions statistically significant at each horizon k.
- Investment efficiency (PIMI):
  - Response of output to public investment shocks significantly stronger in countries with higher public investment efficiency, both short and medium term. Differences economically meaningful but relatively small compared to standard errors; not statistically significant for medium-term growth.

### Robustness checks and diagnostics
- Controlling for growth surprises and growth innovations (difference between actual GDP growth and October forecast) yields results very close to baseline.
  - Correlation between public investment shocks and growth forecast errors: -0.01.
- Adding current and lagged forecast errors of the budget balance produces output response function almost identical to baseline.
- Using alternative forecast vintages (October of previous year, April of same year) gives response functions very close to baseline.
- Sample splits:
  - Emerging markets vs. low-income countries: response of output does not statistically significantly differ between the two groups; effects smaller and less precisely estimated for low-income countries.
- Sensitivity to influential observations:
  - Bootstrapping by randomly dropping 10 percent and 25 percent of countries (100 repetitions): average point estimates across bootstraps very close to baseline; 90 percent confidence bands reported.

### Empirical evidence from tables and figures (highlights)
- Table 1: "The effect of public investment shocks on output" — selected coefficient entries (K=0 to K=5):
  - growth(t-1): 0.139*** (3.13); 0.148** (2.59); 0.139*** (3.13); 0.094 (0.96); -0.006 (0.96); -0.346*** (-2.63)
  - Public investment shock (t): 0.010* (1.73); 0.018*** (2.67); 0.028** (2.33); 0.022** (2.33); 0.041*** (3.00); 0.010* (1.73)
  - Sample sizes N: 1022, 1021, 943, 866, 791, 715 (for K=0 to K=5 respectively)
  - R2: 0.33, 0.37, 0.48, 0.56, 0.64, 0.70
- Figures 1–13 report impulse responses to an unanticipated 10 percent increase in public investment (Figures 1–11, 13) or an unanticipated 10 percent point increase in public consumption (Figure 12) with 90 percent confidence bands. Notation: "t=0 is the year of the shock"; estimates based on equation (2) or equation (3) as indicated.
- Appendix lists sample country coverage and density plots of public investment shocks across subsamples.

### Conclusions and policy implications
- Core empirical conclusions:
  - Public investment increases output in developing economies in both short and medium term.
  - Public investment shocks increase employment in the short term and tend to crowd-in private investment.
  - Fiscal multipliers for public investment (~0.2) are larger than for government consumption (~0.1) in short and medium run.
  - Effects are heterogeneous: larger during periods of slack; in fixed exchange rate regimes; in more closed economies; in countries with lower debt; and in countries with higher investment efficiency.
  - Public investment tends to lower inequality (Gini of market income) over the medium term.
- Policy implications:
  - Public investment can increase output in the short and medium term; to magnify effects, countries should improve investment efficiency and maintain better initial fiscal positions.
  - Shifting expenditure composition from consumption to public investment could be output-enhancing; fiscal consolidations that cut public investment could be particularly costly.
  - Public investment does not appear to pose equity-efficiency tradeoffs and tends to improve both macroeconomic and distributional outcomes.
- Suggested further research:
  - Examine the channels through which increased public investment affects income distribution.

*Source: wp17217 (IMF working paper content provided).*

### References .............................................................................................................

### wp17217 - References

### Introduction and research objective
- Focus: Assess macroeconomic effects of public investment in emerging markets and low-income countries using exogenous identification of unanticipated public investment shocks.
- Sample: unbalanced panel of 79 emerging market and developing economies over the period 1990-2013.
- Motivation: Data limitations preclude natural experiment and SVAR approaches for a large set of countries; the Auerbach and Gorodnichenko (AG) forecast-error approach is employed to mitigate “fiscal foresight” and better align econometrician and agents’ information sets.

### Identification strategy and rationale
- Public investment shocks constructed as forecast errors (FE) using IMF World Economic Outlook October forecasts:
  - FE_{i,t} = ∆lnPI_{i,t} − ∆lnPI^E_{i,t} = (lnPI_{i,t} − lnPI_{i,t−1}) − (lnPI^E_{i,t} − lnPI_{i,t−1})  (equation (1) in source).
- Rationale for October forecast choice: minimizes likelihood that unanticipated within-quarter business cycle news drive FE. The AG approach addresses fiscal foresight and aligns agents’ and econometrician’s information sets.
- Distributional features of FE shocks:
  - Range: -195 to 185 percent.
  - 90 percent of shocks are between -45 and 43 percent.
  - Average (median) shock: about -1.8 (-1.7) percent.
  - Actual growth rate of public investment in full sample ranges from -186 to 7586.

### Empirical method
- Local projection (Jordà, 2005) used to estimate impulse-response functions; estimated for horizons k = 0, . . , 5 (k = 0 is year of shock).
- Baseline specification (equation (2) in source):
  - y_{i,t+k} − y_{i,t−1} = α_i^k + θ_t^k + β^k FE_{i,t} + θ^k X_{i,t} + ε_{i,t}^k
  - y is log output; α_i are country fixed effects; θ_t are time fixed effects; X_it includes two lags of FE and two lags of GDP growth.
  - Confidence bands obtained using clustered robust standard errors at the country level.
- Nonlinear/specification allowing dependence on business cycle and country characteristics (equation (3) in source):
  - y_{i,t+k} − y_{i,t−1} = α_i^k + θ_t^k + β1^k G(z_{it}) FE_{i,t} + β2^k (1−G(z_{it})) FE_{i,t} + θ^k M_{i,t} + ε_{i,t,k}
  - G(z_{it}) = exp(−γ z_{it}) / (1 + exp(−γ z_{it})), γ > 0.
  - z is indicator of business cycle (real GDP growth) or country-specific characteristic normalized to zero mean and unit variance.
  - Public Investment Management Index (PIMI) used as measure of investment efficiency.
  - Parameter setting: γ = 1.5 (results robust to alternative γ).

### Key empirical findings and quantitative results
- Average effects:
  - Public investment increases, on average, output in the short and medium term.
  - Estimated short-term fiscal multiplier for developing economies: about 0.2.
- Heterogeneity: average effects mask differences across episodes and countries; tentative evidence of larger effects when:
  - (i) during periods of slack (business cycle troughs);
  - (ii) in economies operating with fixed exchange rate regimes;
  - (iii) in more closed economies (lower trade openness);
  - (iv) in countries with lower public debt;
  - (v) in countries with higher investment efficiency (as measured by PIMI).
- Distributional effect:
  - Increases in public investment tend to lower income inequality proxied by the Gini coefficient of market income.
- Robustness and diagnostics:
  - Results shown robust to controlling for growth news and when controlling for potential influential observations.
  - Local projection method preferred for flexibility and for estimating nonlinearities; equivalence to STAR modeling noted (Auerbach and Gorodnichenko discussion).

### Contributions to the literature
- Clarifies short- and medium-run causal effects of public investment using forecast-error identification in a large sample of developing economies.
- Complements production-function based long-term elasticity literature by providing identification that mitigates reverse causality and unobservable confounders.
- Extends literature on fiscal policy heterogeneity by examining how business cycle and country-specific factors modulate public investment multipliers.
- Adds to fiscal policy and inequality literature by documenting distributional effects of public investment.

*Source: wp17217 - References (extracted content from the provided PDF).*

### 0.1 percent in the same year. Using the sample average of government investment as

### wp17217 - 0.1 percent in the same year. Using the sample average of government investment as

### Key findings
- Average short-term (public investment spending) fiscal multiplier in the sample of developing economies: about 0.2.
- A 10 percent unanticipated shock to government investment spending:
  - Short-term effect on the level of real output: 0.1 percent in the same year.
  - Five years after the shock, the level of real output is about 0.4 percent higher.
  - Short-term increase in employment: about 0.15 percent.
- Public investment shocks, on average, tend to crowd-in private investment.
- Short-term fiscal multiplier for public consumption: about 0.1 (computed as the ratio of the estimated coefficient to the public consumption-output share), slightly smaller than for public investment (0.2).
- Distributional effect: a 10 percent increase in public investment is associated with a reduction in the Gini coefficient of about 0.2 percent (approximately ¾ of the standard deviation of the average annual change in the Gini coefficient in the sample).
- Long-run / medium-term: public investment delivers larger medium-term output effects than government consumption.

### Comparison with other findings in the literature
- Kraay (2014) finds an average fiscal multiplier ranging between 0.2 and 0.4 for a sample of 109 developing economies.
- Ilzetzki and others (2013) find a fiscal multiplier between 0 and 0.6 for a panel of 24 emerging market economies.
- Abiad and others (2016) find the short-term multiplier associated with public investment shock in advanced economies is about 0.5.
- The smaller effects in developing economies may reflect lower investment process efficiency (Pritchett 2000), which is examined empirically.

### Robustness checks
- Controlling for current and lagged output growth innovations (difference between actual GDP growth and the rate forecast by analysts in October of the same year) yields results very close to the baseline.
  - Correlation between public investment shocks and growth forecast errors: -0.01.
- Adding current and lagged forecast errors of the budget balance produces a response function of output almost identical to the baseline.
- Using the forecasts of the October issue of the previous year (instead of October of the same year) to compute government investment forecast errors yields a response function of real output very close to baseline.
  - Similar results obtained using forecasts of the April issue of the same year.
- Splitting the sample by estimation periods and country samples:
  - Estimating separately for emerging markets and low-income countries: response of output does not statistically significantly differ between the two groups, though effects are smaller and less precisely estimated for low-income countries.
- Sensitivity to influential observations:
  - Re-estimating equation (2) by randomly dropping 10 percent and 25 percent of countries, repeated 100 times (bootstraps): average point estimates across bootstraps are very close to and not statistically different from the baseline; 90 percent confidence bands reported.

### Effects across episodes- and country-specific characteristics
- Role of the business cycle:
  - Response of output tends to be higher during periods of recessions than in expansions; estimated differences are small relative to standard errors and mostly not statistically significant.
  - The difference is statistically significant only 4 years after the public investment shock.
- Exchange rate regime:
  - Effects substantially stronger in countries with fixed exchange rate regimes: increase in GDP positive under fixed regimes, negative (and insignificant) under flexible regimes.
  - Given large standard errors, difference in output response is statistically significant only for k=0 and k=4.
- Trade openness:
  - Output response positive in countries with very low trade openness, negative (and insignificant) in countries with very high trade openness.
  - Differences in point estimates small relative to standard errors and mostly not statistically significant.
- Debt-to-GDP:
  - Effect of public investment shocks significantly varies with initial public debt positions:
    - Countries with very low debt: response of output positive, significant and higher than baseline.
    - Countries with very high debt: response negative and not statistically significant.
  - Difference in impulse response functions is statistically significant at each horizon k.
- Investment efficiency (PIMI indicator):
  - Response of output to public investment shocks is significantly stronger in countries with higher public investment efficiency, both short and medium term.
  - Differences economically significant but relatively small compared to standard errors; result not statistically significant for medium-term growth.
  - PIMI index captures institutional environment across project appraisal, selection, implementation, and evaluation stages.

### Public investment versus government consumption
- Public consumption effect on output significant only in the short run (at the time of the shock).
- Short-term fiscal multiplier:
  - Public investment: about 0.2.
  - Public consumption: about 0.1.
- Evidence indicates public investment has a larger fiscal multiplier than government consumption, consistent with Ilzetzki and others (2013).

### Distributional effects
- Replacing output with the Gini coefficient of market income in estimation:
  - Public investment increases tend to reduce inequality in short and medium term.
  - Quantified effect: 10 percent increase in public investment → reduction in Gini by about 0.2 percent.

### Conclusions and policy implications
- Public investment in infrastructure is an important driver of economic development in emerging market and low-income countries.
- Key empirical conclusions:
  - Public investment increases output in developing economies in both short and medium term.
  - Public investment shocks increase employment in the short term and tend to crowd-in private investment.
  - Fiscal multipliers associated with public investment are larger than those for government consumption in short and medium run.
  - Effects are heterogeneous: larger during periods of slack; in fixed exchange rate regimes; in more closed economies; in countries with lower debt; and in countries with higher investment efficiency.
  - Public investment tends to lower inequality (Gini of market income) over the medium term.
- Policy implications:
  - Public investment has potential to increase output in short and medium term; to magnify effects, countries should improve investment efficiency and have better initial fiscal positions.
  - Shifting composition of government expenditure away from consumption and toward public investment could be output-enhancing; conversely, fiscal consolidations based on public investment cuts could be particularly painful.
  - Public investment does not seem to pose equity-efficiency tradeoffs and tends to improve both macroeconomic and distributional outcomes.
- Broader context:
  - McKinsey Global Institute estimate: from 2016 through 2030, the world needs to invest about 3.8 percent of GDP—or an average of $3.3 trillion a year—in economic infrastructure to support expected growth; developing economies account for some 60 percent of that need.
- Suggested avenue for further research:
  - Examine the channels through which increased public investment affects income distribution.

*Source: IMF Working Paper content (wp17217).*

### REFERENCES

### REFERENCES

### Cited literature
- Abiad, A., D. Furceri and P. Topalova. “The Macroeconomic Effects of Public Investment: Evidence from Advanced Economies.” Journal of Macroeconomics, 50, pages 224-240, 2016.
- Adam, C., and D. Bevan. “Aid and the Supply Side: Public Investment, Export Performance, and Dutch Disease in Low-Income Countries.” The World Bank Economic Review, vol. 20, no. 2, Apr. 2006, pp. 261–290.
- Auerbach, Alan J, and Yuriy Gorodnichenko. “Measuring the Output Responses to Fiscal Policy.” American Economic Journal: Economic Policy, vol. 4, no. 2, pp. 1–27, 2012a.
- Auerbach, Alan J, and Yuriy Gorodnichenko. “Measuring the Output Responses to Fiscal Policy.” "Fiscal Multipliers in Recession and Expansion," NBER Chapters, in: Fiscal Policy after the Financial Crisis, pages 63-98 National Bureau of Economic Research, Inc. 2012b.
- Auerbach, Alan J, and Yuriy Gorodnichenko. “Output Spillovers from Fiscal Policy,” American Economic Review, vol. 103(3), pages 141-146, 2013.
- Barro, Robert. “Government Spending in a Simple Model of Endogenous Growth.” Journal of Political Economy, vol. 95, 1990, pp. 103–126.
- Barro, Robert., and Xavier Sala-I-Martin. “Public Finance in Models of Economic Growth.” The Review of Economic Studies, vol. 59, no. 4, 1992, pp. 645–661.
- Berg, A. Buffie, E. Pattillo, C., Portillo, R. Presbitero, A. and L. Zanna, 2015. "Some Misconceptions about Public Investment Efficiency and Growth," IMF Working Papers 15/272, International Monetary Fund.
- Blanchard, O., and R. Perotti. “An Empirical Characterization of the Dynamic Effects of Changes in Government Spending and Taxes on Output.” The Quarterly Journal of Economics, vol. 117, no. 4, Jan. 2002, pp. 1329–1368.
- Blanchard, Olivier, and Daniel Leigh. “Growth Forecast Errors and Fiscal Multipliers.” American Economic Review, vol. 103, no. 3, 2013, pp. 117–120.
- Bom, Pedro R., and Jenny E. Ligthart. “What Have We Learned From Three Decades Of Research On The Productivity Of Public Capital?” Journal of Economic Surveys, vol. 28, no. 5, 2014, pp. 889–916.
- Calderón, César, and Luis Servén. “The Effects of Infrastructure Development on Growth and Income Distribution.” World Bank Policy Research Working Papers, 2004.
- Calderon, César, and L. Serven. “Infrastructure and Economic Development in Sub-Saharan Africa.” Journal of African Economies, vol. 19, no. 1, 2010, pp. 13–87.
- Chatterjee, Santanu, and Stephen J. Turnovsky. “Foreign Aid and Economic Growth: The Role of Flexible Labor Supply.” Journal of Development Economics, vol. 84, no. 1, 2007, pp. 507–533.
- Corsetti, Giancarlo; Meier, André; Müller, Gernot J., 2012. "What determines government spending multipliers?," Economic Policy, vol. 27(72), pages 521-565, October.
- Eden, Maya, and Aart Kraay. “‘Crowding in’ and the Returns to Government Investment in Low-Income Countries.” Policy Research Working Papers, 2014.
- Favero, Carlo, and Francesco Giavazzi. “How Large Are the Effects of Tax Changes?” NBER Working Paper, 2009.
- Forni, Mario, and Luca Gambetti. “Fiscal Foresight and the Effects of Government Spending.” CEPR Discussion Paper, 2010.
- Futagami, Koichi et al. “Dynamic Analysis of an Endogenous Growth Model with Public Capital.” The Scandinavian Journal of Economics, vol. 95, no. 4, 1993, pp. 607–625.
- Glomm, Gerhard, and B. Ravikumar. “Public Investment in Infrastructure in a Simple Growth Model.” Journal of Economic Dynamics and Control, vol. 18, no. 6, 1994, pp. 1173–1187.
- Granger, C. W. J., and Teräsvirta Timo. Modelling Nonlinear Economic Relationships. Oxford, Oxford University Press, 1993.
- Huidrom, R, M A Kose, J J Lim, and F L Ohnsorge (2016b), “Do fiscal multipliers depend on fiscal positions?”, Policy Research Working Paper Series 7734, The World Bank.
- Ilzetzki, E., Mendoza E. G., and Végh C. A. “How Big (Small?) are Fiscal Multipliers?” Journal of Monetary Economics, 60:2, 2013 pp. 239-254.
- Jordà, Òscar. “Estimation and Inference of Impulse Responses by Local Projections.” American Economic Review, vol. 95, no. 1, 2005, pp. 161–182.
- Kraay, A. “How Large Is the Government Spending Multiplier? Evidence from World Bank Lending.” The Quarterly Journal of Economics, vol. 127, no. 2, 2012, pp. 829–887.
- Kraay, Aart. “Government Spending Multipliers in Developing Countries: Evidence from Lending by Official Creditors.” American Economic Journal: Macroeconomics, vol. 6, no. 4, 2014, pp. 170–208.
- Leeper, Eric M et al. “Corrigendum: Quantitative Effects of Fiscal Foresight.” American Economic Journal: Economic Policy, vol. 4, no. 3, 2012, pp. 283–283.
- Leeper, Eric M. et al. “Fiscal Foresight and Information Flows.” Econometrica, vol. 81, no. 3, 2013, pp. 1115–1145.
- Portillo, Rafael et al. “The Macroeconomics of Medium-Term Aid Scaling-Up Scenarios.” IMF Working Papers, vol. 10, no. 160, 2010.
- Portillo, Rafael et al. “Public Investment, Growth, and Debt Sustainability: Putting Together the Pieces.” IMF Working Papers, vol. 12, no. 144, 2012.
- Romer, Christina D, and David H Romer. “The Macroeconomic Effects of Tax Changes: Estimates Based on a New Measure of Fiscal Shocks.” American Economic Review, vol. 100, no. 3, 2010, pp. 763–801.
- Romp, Ward, and Jakob De Haan. “Public Capital and Economic Growth: A Critical Survey.” Perspektiven Der Wirtschaftspolitik, vol. 8, no. S1, 2007, pp. 6–52.
- Stock, James H., and Mark W. Watson. “Why Has U.S. Inflation Become Harder to Forecast?” Journal of Money, Credit and Banking, vol. 39, 2007, pp. 3–33.
- Zeev, Nadav Ben, and Evi Pappa. “Chronicle of a War Foretold: The Macroeconomic Effects of Anticipated Defense Spending Shocks.” The Economic Journal, 2015.

### Figures, tables, and empirical specifications (content highlights)
- Figures (1–13) report impulse responses to an unanticipated 10 percent increase in public investment (Figures 1–11) or an unanticipated 10 percent point increase in public consumption (Figure 12), with dashed lines denoting 90 percent confidence bands. Notes repeatedly state: "t=0 is the year of the shock" and "Estimates based on equation (2)" or "Estimates based on equation (3)" as indicated.
- Figure 1: "The effect of public investment shocks on output (percent)". Solid blue lines denote the response to an unanticipated 10 percent increase in public investment.
- Figure 2: "The effect of public investment shocks on employment (percent)". Solid blue lines denote the response to an unanticipated 10 percent increase in public investment.
- Figure 3: "The effect of public investment shocks on private investment (percent)". Solid blue lines denote the response to an unanticipated 10 percent increase in public investment.
- Robustness and heterogeneity checks across Figures 4–11:
  - Figure 4: Robustness checks I — Panels A–D: controlling for growth surprises; controlling for budget surprises; using previous year forecast; post 2005. Yellow lines are baseline effects reported in Figure 1.
  - Figure 5: Emerging markets vs. low-income countries. Classification follows IMF World Economic Outlook.
  - Figure 6: Controlling for influential observations — Panel A: Dropping 10% of countries at a time; Panel B: Dropping 25% of countries at a time. Results shown as averages from 100 replications.
  - Figure 7: The role of business cycle — Panel A: Low growth; Panel B: High growth. Effects under strong (weak) business cycle conditions are shown using G=0 (G=1). Estimates based on equation (3).
  - Figure 8: The role of exchange rate regimes — Panel A: Fixed exchange rate regime; Panel B: Flexible exchange rate regime. Effects shown using G=0 (G=1). Estimates based on equation (3).
  - Figure 9: The role of trade openness — Panel A: Low trade openness; Panel B: High trade openness. Effects shown using G=0 (G=1). Estimates based on equation (3).
  - Figure 10: The role of public debt — Panel A: Low public debt-to-GDP ratio; Panel B: High public debt-to-GDP ratio. Effects shown using G=0 (G=1). Estimates based on equation (3).
  - Figure 11: The role of investment efficiency — Panel A: High investment efficiency; Panel B: Low investment efficiency. Effects shown using G=0 (G=1). Estimates based on equation (3).
- Figure 12: "The effect of public consumption shocks on output (percent)". Solid blue lines denote the response to an unanticipated 10 percent point of increase in public consumption. Estimates based on equation (2).
- Figure 13: "The effect of public investment shocks on inequality (percent)". Solid blue lines denote the response to an unanticipated 10 percent point of increase in public investment. Estimates based on equation (2).
- Table 1: "The effect of public investment shocks on output"
  - Columns K=0 to K=5 report coefficient estimates and t-statistics (in parentheses). Selected entries:
    - growth(t-1): 0.139*** (3.13); 0.148** (2.59); 0.139*** (3.13); 0.094 (0.96); -0.006 (0.96); -0.346*** (-2.63)
    - growth(t-2): -0.002 (-0.06); 0.056 (0.47); -0.002 (-0.06); -0.080 (-0.49); -0.215 (-0.93); -0.134 (-0.48)
    - Public investment shock (t): 0.010* (1.73); 0.018*** (2.67); 0.028** (2.33); 0.022** (2.33); 0.041*** (3.00); 0.010* (1.73)
    - Public investment shock (t-1): 0.001 (1.07); -0.000 (-0.07); 0.000 (0.35); 0.000 (0.48); 0.000 (0.55); 0.010* (1.65)
    - Public investment shock (t-2): -0.001 (-1.13); -0.001 (-0.60); -0.000 (-0.03); -0.000 (-0.32); 0.001 (0.95); -0.000 (-0.37)
  - Sample sizes and fit:
    - N: 1022, 1021, 943, 866, 791, 715 (for K=0 to K=5 respectively)
    - R2: 0.33, 0.37, 0.48, 0.56, 0.64, 0.70
  - Note: T-statistics based on robust clustered standard errors in parenthesis. ***,**, * denote significance at 1 percent, 5 percent and 10 percent, respectively. Estimates based on equation (3).
- Appendix material:
  - Figure A1: "Public investment shocks (percent)" — density plots for Panel A: All sample; Panel B: Emerging markets; Panel C: Low-income countries. Note: classification follows IMF WEO.
  - Table A1: "List of countries in the sample" — binary indicators for Emerging markets (1/0) and Low-income (1/0). Examples from the table:
    - Afghanistan 0 1
    - Algeria 1 0
    - Angola 1 0
    - Armenia 1 0
    - Bangladesh 0 1
    - Benin 0 1
    - Brazil 1 0
    - Burkina Faso 0 1
    - Cameroon 0 1
    - China 1 0
    - ... (table continues with full sample list)
  - Table A2: "List of countries and the sample periods" — country coverages (examples):
    - Afghanistan 2005-2013
    - Algeria 1995-2013
    - Angola 2006-2013
    - Armenia 2002-2013
    - Bahrain 1996-2013
    - Bangladesh 1990-2013
    - Barbados 1995-2013
    - Benin 2001-2013
    - Bosnia Herzegovina 2003-2013
    - Botswana 2001-2013
    - Brazil 1991-2013
    - Burkina Faso 2001-2013
    - Cameroon 2001-2013
    - Capo Verde 2003-2013
    - Central African Rep. 2001-2013
    - Chad 2006-2013
    - China 2003-2013
    - Comoros 2001-2013
    - Congo, Republic of 2001-2013
    - Costa Rica 2001-2013
    - Cote d'Ivoire 2001-2013
    - Djibouti 1996-2013
    - Dominican Republic 1998-2013
    - Egypt 1991-2013
    - El Salvador 1992-2013
    - Equatorial Guinea 2001-2013
    - Eritrea 2001-2013
    - Ethiopia 2001-2013
    - Gabon 1992-2013
    - Gambia, the 2001-2013
    - Ghana 1991-2013
    - Guatemala 2002-2013
    - Guinea 1997-2013
    - Guinea-Bissau 2004-2013
    - Haiti 1996-2013
    - Honduras 2001-2013
    - Iran 1992-2013
    - Jordan 1991-2013
    - Kuwait 1992-2013
    - Lebanon 1998-2013
    - Lesotho 2001-2013
    - Libya 2000-2013
    - Madagascar 2001-2013
    - Malawi 2006-2013
    - Mali 2001-2013
    - Mauritania 1999-2013
    - Mauritius 2001-2013
    - Mexico 1990-2013
    - Moldova 1998-2013
    - Morocco 1999-2013
    - Mozambique 2001-2013
    - Namibia 2001-2013
    - Nicaragua 2002-2013
    - Nigeria 2000-2013
    - Oman 1993-2013
    - Pakistan 1992-2013
    - Peru 1991-2013
    - Qatar 2000-2013
    - Romania 2001-2013
    - Rwanda 2001-2013
    - Sao Tome & Pr. 2002-2013
    - Saudi Arabia 1995-2013
    - Senegal 2001-2013
    - Serbia 2007-2013
    - Seychelles 2007-2013
    - Sierra Leone 2004-2013
    - South Africa 2001-2013
    - Swaziland 2001-2013
    - Tanzania 2006-2013
    - Thailand 1991-2013
    - Togo 2001-2013
    - Turkey 1990-2013
    - Uganda 2001-2013
    - United Arab Em. 1994-2013
    - Uruguay 1998-2013
    - Venezuela 1998-2013
    - Yemen 1999-2013
    - Zambia 2001-2013
    - Zimbabwe 2012-2013

*Source: wp17217 - REFERENCES (IMF working paper content as provided).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp17217.pdf_